NEW PROSPECTS FOR LATIN AMERICAN BONDS MARKET AFTER … · NEW PROSPECTS FOR LATIN AMERICAN BONDS MARKET AFTER ... The motivation for the present thesis is related to the evaluation

A. Díaz Montenegro / Erasmus School of Economics

1

POSITIVE WAKE–UP–CALL CONTAGION:

NEW PROSPECTS FOR LATIN AMERICAN BONDS MARKET AFTER THE GREEK

DEBT CRISIS

ERASMUS UNIVERSITY ROTTERDAM

Erasmus School of Economics

Department of Economics

Supervisor: Barbara Sadaba

Author: Alexandra Díaz Montenegro

E-mail address: [email protected]


2

Abstract

Recently, Latin American countries have experienced a series of upgrades in their sovereign

credit ratings that are reflecting the region’s appeal for sovereign portfolio managers around the

globe. Paradoxically, Latin American economies have shown strong economic fundamentals

during the whole last decade and did not experienced any significant local common shock during

2009 - 2012, years in which credit agencies started upgrading the ratings of the region. Given

these facts a valid question that springs up is which circumstances triggered the change in

perception over the region’s creditworthiness during the last years. In this document I intend to

explore a possible explanation to this change in risk perception, which consists in a positive

wake–up-call contagion generated after the 2009 Greek debt crisis. The thesis will follow the

methodology proposed in Giordano, Pericolli and Tomassino (2012), in which an analysis of

contagion was conducted exclusively over developed European countries. Estimations show

robust evidence of positive wake – up – call contagion towards the Latin American region. This

outcome confirms that pricing of Latin American sovereign bonds in international financial

markets is finally acknowledging the efforts of Latin American economies in correcting fiscal

imbalances, implementing inflation targeting regimes and creating capital buffers to counter the

effects of new coming local and international crises.


3

I. Introduction

Recently, Latin American countries have experienced a series of upgrades in their sovereign

credit ratings that are reflecting the region’s appeal for sovereign portfolio managers around the

globe. For instance, Mexico’s creditworthiness qualification was upgraded in 2013 by Standard

& Poors (S&P), Colombia got the investment grade from S&P, Fitch and Moody’s in 2011,

Brazil’s profile was upgraded in September 2009 by Moody’s and in 2008 by S&P, and Chile

reached a AA- level in its sovereign rating in 2012 (Reuters, Scotiabank, Bakermckenzie and

Bloomberg).

Indeed, even though the region has not been completely immune to the recent world financial

disasters of the 2008 subprime crisis and the Eurozone debt crisis (Cárdenas and Henao (2010),

Ocampo (2009)), the main credit agencies have bet positively in the region’s probability of

default. Paradoxically, Latin American economies have shown strong economic fundamentals

during the whole last decade (International Monetary Fund – Regional Economic Report.

Western Hemisphere. Time to Rebuild Policy Space May 2013) and did not experienced any

significant local common shock during 2009 - 2012, years in which credit agencies started

upgrading the ratings of the region. Given these facts a valid question that springs up is which

circumstances triggered this behavior; what is the cause of the change in perception over the

region’s creditworthiness during the last years.

In this document I intend to explore a possible answer which consists in a positive wake–up-call

contagion generated after the 2009 Greek debt crisis. My hypothesis lays on the presumption that

the Greek bailout not only led to a reassessment of the fundamentals of other European

economies, as suggested in Giordano, Pericoli, & Tommasino (2012), but also led to a

reassessment of the fundamentals of emerging economies such as the ones located in the Latin

American region.

This document contributes to the literature in two ways. The first contribution is in proposing an

analysis of a positive rather than a negative wake - up - call contagion generated after the Greek


4

debt crisis. Usually, contagion is analyzed with a negative connotation; nevertheless, crises in a

region can actually represent an opportunity for greenfield investments or the deepening of

incipient existing investments in different promising countries. The latter represents a small but

significant deviation from the analysis proposed in Giordano et al. (2012), the paper that first

introduced the analysis of wake –up – call episodes within the European economies after the

2008 Greek distress. The second contribution to the literature is the especial focus on Latin

American countries, nations that belong to a new promising region in which the strongest

economies have enhanced their fiscal positions, implemented inflation targeting regimes and

have created countercyclical mechanisms in order to avoid the effects of crises.

The rest of the document is organized as follows, chapter II presents a relevant literature review,

chapter III explains the methodology used to tackle the research question, chapter IV presents the

data collection process and results, and finally, chapter V concludes and presents further

discussion points on the topic.

II. Literature Review

The motivation for the present thesis is related to the evaluation of the determinants of yield

spreads before and after the Greek Debt crisis, analyzing whether there has been a significant

change in the market’s assessment of macroeconomic fundamentals in the Latin American

region. In the next paragraphs I will highlight how the existing literature has addressed the topic,

which models have been used, the choice of variables and finally, how contagion behavior has

been analyzed.

Indeed, the assessment of determinants of sovereign bond yield spreads has been a vastly studied

subject among academics. Theoretical models presented in the literature argue that foreign

funding for small and open economies comprise two elements: a risk free world interest rate and

a country risk premium, the latter being a function of the country’s probability of default. The

determinants of the probability of default are related to solvency and liquidity variables,

characteristics that show whether an economy is capable of servicing its debt. Such framework is


5

proposed in Edwards (1986), one of the seminal papers on the topic that leaded the research by

extending the analysis of risk default premium of international bank loans to the government

bonds market. His main findings provide support for the positive relationship between high debt

ratios and high risk premiums in emerging markets.

Since then and despite the little theoretical guidance about which specific variables should be

included in an empirical model of risk premium (Ebner (2009)), a large number of studies have

assessed different explanatory variables as determinants of yield spreads. The variables can be

grouped, as proposed in Min (1998), in four categories: a) macroeconomic fundamentals, b)

liquidity and solvency variables, c) external market conditions and shocks, and d) dummy

variables. Regarding the first category, measures of economic soundness have been evaluated

using proxy variables such as investment to GNP ratio, imports to GDP ratio, index of real

effective exchange rate, volatility of terms of trade and inflation rate (Edwards (1986), Baldacci,

Mati & Gupta (2008), Bellas, Papaioannou & Petrova (2010), Eichengreen & Mody (1998), Min

(1998), Hilscher & Nosbusch (2010)). These variables capture the long term capacity of a

country to repay its liabilities. Meanwhile, liquidity and solvency conditions have been

evaluated including variables such as the GDP growth, international reserves to GNP ratio,

current account to GNP ratio, volatility of terms of trade, debt service ratio, inflation rate,

consumer price index and industrial production level (Edwards (1986), Sachs (1981), Baldacci et

al. (2008), Ebner (2009)). Additionally, market conditions or the so called global risk appetite,

have also been incorporated using proxies such as the VIX volatility index1, Moody’s index,

S&P index, the US policy interest rate and the VDAX-NEW, which measures the implicit

volatility of the German stock index DAX over a period of 30 days (Hilscher & Nosbusch

(2010), Baldacciet al. (2008), Mody (2009), Kamin & Von Kleist (1999), Ebner (2009)). Other

types of variables used, especially in emerging markets literature, include the country’s history

of default, interest rates of different maturities and political instability indexes (Hilscher &

Nosbusch (2010), Reinhart, Rogoff & Savastano (2003), Bellas et al. (2010), Eichengreen &

Mody (1998), Min (1998), Baldacci et al. (2008)).

1 VIX is the short name of The Chicago Board Options Exchange Market Volatility Index. According to Giordano et

al. (2012) VIX is regarded as a good indicator of the level of fear or greed in U.S. and global capital markets.


6

Recently, the literature has given particular attention to the behavior of yield spreads before and

after the subprime crisis of 2007 and the European debt crisis of 2008. Caceres, Guzzo and

Segoviano (2010) evaluate whether the fluctuations in yield spreads of the European economies

after 2007 are caused by global risk aversion, changes in fundamentals or contagion from other

countries. Their results show that the behaviour of fundamentals was essentially the driven force

of the movements of swap yield spreads after the crisis. In line with this result Constâncio (2012)

finds that contagion effects of the Greek debt crisis accounted for an increase of 25 to 45% in the

yield spreads in Portugal and Ireland. Meanwhile, Sgherri and Zoli (2009) and Arghyrou and

Kontonikas (2012) show that after 2007/08, markets “enlarged the weight” given to national

fiscal variables and expectations of debt performance as determinants of yield spreads. As a

consequence, nowadays markets are differentiating more between bonds from strong economies

and weaker ones within the European Union.

In line with these results, Giordano et al. (2012) find evidence of a specific type of contagion, a

wake – up – call contagion or reassessment of the fundamentals of other European economies

after the Greek debt crisis. The authors analyse three types of contagion in the Eurozone after the

crisis, a wake-up-call contagion, shift contagion and pure contagion. Wake-up-call contagion

refers to a situation in which fundamentals of a specific country are not priced correctly but an

episode of crisis in another country triggers a change in investor’s risk perception via a

reevaluation of its fundamentals. The study stresses that under wake-up-call contagion a crisis

that takes place in one country provides new information such that investors reevaluate their risk

perception towards other countries. Shift contagion is a situation where the risk perception of a

country changes given an increased sensitivity to common factors such as global risk aversion.

Finally, pure contagion refers to a complete loss of confidence, irrational herding behavior and

margin calls. The authors do not find evidence of the presence of these two last types of

contagion.

As it is clear from the description above, much attention has been paid to the effects of the Greek

debt crisis on the spreads in the Eurozone. However, less consideration has been given to the

potential collateral effects on emerging and developing economies. On the one hand, according

to the Global Financial Stability Report (Market Update - January 2012) issued by the


7

International Monetary Fund and Ocampo (2009), global financial crises can have several

negative effects on the emerging economies, for instance, these economies can expect a decrease

in credit, a deterioration of business climate, a decrease in foreign direct investment, a decrease

in exports and a decrease in developing aid. On the other hand, the IMF’s report also states that

many of these economies have saved enough capital that allows them to counter the negative

effects of international shocks and implement countercyclical policies to avoid any damage to

their internal economies. In other words, emerging economies have learnt from past chapters of

distress, have created strong institutions and have built more stable and resistant economies to

local as well as to external crises.

Indeed, according to Resende & Goldfajn (2012) and Montoro y Rojas-Suarez (2012), the Latin

American is one of the regions that has been most resilient to the 2008 financial crisis given their

good fundamentals, their good external positions, their successful system of inflation targeting,

their liquidity buffers created after the local financial crises experienced during the 90s, and

given an increase in the demand from China. The authors also find statistical evidence that

supports the hypothesis of Latin America as a less exposed region to external shocks. In line with

these findings Goldberg (2005) finds that the relationship between U.S. bank claims to Latin

American economies is not linked to the U.S. business cycle. The later signifies that crises

originated in this economy are unlikely to spread to the Latin American region. Notwithstanding,

Galindo, Izquierdo & Rojas-Suarez (2010) state that efforts to create a more financially

integrated system between the Latin American region and international financial markets also

created a channel through which the 2008 crisis spread very quickly. Indeed, after 1990 Latin

American economies engaged in more liberal policies that included international financial

integration. As a consequence they experienced an increase in international financial transactions

and an increased number of international financial institutions such as banks with presence in the

countries. The authors explain that the existence of foreign banks in countries such as the Latin

Americans represents in one hand, an alternative option that diversifies risk for the local

economy, but on the other hand it represents also a gate through which international crises can

spread easily. The diversification of risk would be accomplished given the capital buffers owned

by the parent bank which could rescue the local subsidiary in case of an internal shock. In turn,

in case of an external shock foreign banks can reduce cross border lending immediately and they


8

can also reduce the subsidiary’s lending in the local economy. The authors find statistical

evidence suggesting that the presence of foreign banks amplifies external financial shocks in the

local economies, thus, Latin American economies experienced a case of negative contagion after

the subprime crisis of 2008. A more recent article, Martinez and Ramirez (2011), analyze the

existence of contagion in the equity markets of some Latin American economies during the

period of 1990 and 2008. Using a GARCH model and principal components, the authors find no

evidence of financial contagion but the existence of an interdependent relationship across

markets. This result means that the local markets reacted smoothly and linked to the evolution of

the country’s fundamentals. Regarding Government Bonds Markets, which is the main topic of

interest in the present thesis, Jara, Moreno, and Tovar (2009) argue that Latin America reacted

positively after the 2008 financial crisis and sovereign bonds served as a “spare tire” substituting

the cutback in foreign currency lending that followed the burst of the financial crisis.

All in all, the existing literature shows nothing but mixed results. Therefore, the question

whether the European sovereign debt crisis had an effect on the Latin American region is still

open for debate. Knowing whether there has been a negative impact through a more financially

integrated system or a positive reassessment of the better economic position of these economies

is vital for future investment decision towards the region and vital for local policy makers in their

efforts to attract international financial investors. In consequence, as mentioned in the

introductory section of the present document, this paper will focus on the following questions:

Has the Greek debt crisis led to a positive reassessment of the fundamentals of non-European

developing economies such as the Latin American? Using Giordano et al. (2012) argot the

question can be rewritten as follows: Is there evidence of a positive wake–up-call contagion from

the Greek debt crisis to the Latin American region?

III. Methodology & Empirical Specification

As mention previously Giordano et al. (2012) provide evidence of a wake–up-call contagion

from the Greek debt crisis to other European economies in distress. The authors argue that the


9

burst in the Greek financial accounts made investors more fearful towards other European

economies with similar fundamentals creating a negative contagion towards the region.

In order to analyze the case for the Latin American economies, this thesis will follow the

methodology used by Giordano et al. (2012) and will enlarge the sample of countries to include

the major economies of Latin America. This way the analysis will focus on the following

European economies: Austria, Belgium, Finland, France, Ireland, Italy, Portugal, Spain and The

Netherlands; and will also add the major economies in the Latin American region: Argentina,

Brazil, Chile, Colombia, Mexico, Peru and Venezuela. These economies account for 92% of the

total GDP of Latin American and the Caribbean region (Cardenas & Henao (2010)), thus, it can

be considered a representative sample of the region.

Giordano et al. (2012) implements a frequently used specification were the determinants of yield

spreads are grouped in two vectors, one of country specific factors (Z) – including

macroeconomic fundamentals, liquidity and solvency variables – and another vector containing

common factors (F)2:

(1)

| | ,

where denotes the risk of default or yield spread, is a vector of country specific variables,

is a vector of variables which are common across countries and is the error term.

In order to analyze the three types of contagion explained in chapter No. 2 - wake-up-call

contagion, shift contagion and pure contagion – the authors estimate a broader model:

2 This categorization is compatible with the one proposed by Min (1998) and explained in the literature review. For

simplification purposes the macroeconomic fundamentals and liquidity and solvency variables were grouped in one

single vector. This allows a smoother analysis on contagion, which is the main purpose of this document.


10

(2)

| | | | ,

where still denotes yield spread, is the vector of country specific variables, is the vector

of variables which are common across countries and is a dummy variable taking the number

of 1 since October 2009, date in which the Greek Government announced an increase on its debt

to GDP ratio from 7.7 percent to 12.7 percent (Giordano et al. (2012) and Arghyrou and

Tsoukalas (2011)).

Note that wake-up-call contagion is captured by , the coefficients related to the interacted

terms between the fundamentals and the dummy variable. This coefficient captures the extent to

which the effect of fundamentals over the yield spread changes after the Greek Debt crisis. Shift

contagion is captured by the coefficient that accompanies the interacted terms between the

dummy and common factors; this coefficient captures the extent to which the effect common

cross border characteristics influencing the risk premium changes after the Greek Debt crisis

episode. Finally, pure contagion is captured by the coefficient of the dummy variable itself.

In order to analyse whether there has been a positive wake-up-call contagion in the Latin

American Economies, equation (2) is calculated using the fixed effects estimator for the

complete panel and also for a panel using only Latin American countries. Giordano et al. (2012)

use the Least Squares Dummy Variable (LSDV) technique, which controls for country time

invariant effects; nevertheless, the same estimator for the coefficients can be obtained using the

fixed effects estimator which avoids the unpleasantness of calculating a coefficient for each

country dummy variable that LSDV technique requires (Verbeek (2012) Pag.377). The technique

is appropriate for the purpose of this paper since it allows analysing to what extent differs

from ̅, differences within the countries (Verbeek (2012) Pag.379).

The literature warns about the typical presence of heteroskedasticity and autocorrelation in these

types of equations. For this reason equation (2) will be estimated using Heteroskedasticity and

Autocorrelation Consistent Standard Errors (HAC) or robust standard errors, which are more


11

accurate when the form of heteroskedasticity and autocorrelation is unknown (Verbeek (2012)

Pag.103). Problems of dynamic inconsistency that generates overestimated coefficients, is

expected not to be an obstacle in this particular estimation. Verbeek (2012 Pag. 397), shows that

this endogeneity noise represents a major obstacle when both the number of cross sections and

time period are small. However, he also states that a short sample period is consider to be T=10,

which is not the case for the estimations presented in this thesis. Finally, a number of robustness

checks are also conducted in order to analyse the stability of the results. These checks include

instrumental variables and an alternative dummy crisis variable that takes into consideration the

effects of the subprime crisis of 2008.

IV. Data & Results

Data

In line with Giordano et al. (2012) the analysis considers monthly observations of the 17

countries in the sample – 17 European and 7 Latin American - for the period of January 2000 to

December 2012. This accounts for a total of 2652 observations for each one of the variables. The

specific variables that were used are listed in Table No. 1:

Table No. 1

Variable Description

Yield Spread Yield spread of the 10 year sovereign reference bond

* Government Debt Debt/GDP: a measure of liquidity (Edwards (1986)), or a macroeconomic

fundamental (Giordano et al. (2012)).

Private Debt Private Debt / GDP: a measure of the degree of domestic leverage of a

country.

GDP growth An indicator of a economy's soundness and prospects of dynamism

Current account surplus Current Account / GDP: a measure of the degree of external leverage of a

country.

Liquidity Bid-Ask spread: a measure of liquidity, a small gap signals a liquid

market.

VIX Propensity of investors to bear the credit risk

*Following Giordano et al. (2012), macroeconomic variables are differenced with respect to its corresponding

benchmark.


12

The independent variable was built upon the yield of the 10 years maturity sovereign reference

bond with respect to Germany and the US, for the European and Latin American countries

respectively3. Macroeconomic fundamentals considered in Giordano et al. (2012) are the GDP

growth and the current account balance as measures of soundness, dynamism and capacity of a

country to repay its debts; the level of debt to GDP is considered in the model as a measure of

indebtness, a high indebted economy is usually associated with instability and an increased

probability of defaulting in its financial obligations; and finally, the private debt ratio to GDP is

included as a measure of domestic leverage.

Information of macroeconomic variables was obtained from different sources including The

Interamerican Development Bank, Eurostats, OECD and The World Bank. In most of the cases

macroeconomic data is issued quarterly, for this reason a transformation was performed in order

to obtain the monthly series; following Giordano et al. (2012) quarterly observations were left

constant within the quarter. For financial variables such as Bid-Ask Spread and the VIX, data

was taken from Bloomberg and was found published in a monthly basis. The Bid – Ask Spread

was calculated using information available for the Bid Price and Ask Price of the 10 years

maturity sovereign reference bond for each country, except for Argentina since this is a security

that is not reported in Bloomberg’s data base. Regarding the dependent variable, monthly

information of the Mid Yield of the 10 years maturity sovereign reference bond was taken from

Bloomberg. Missing values were completed with information available in Thomson Reuters

Government Bid Yield 10Y Index, which provides information of the local currency reference

bond of each month. Data for Argentina is entirely taken from TR, meanwhile, 65 observations

were completed in the case of Brazil, 50 in the case of Chile, 78 in the case of Colombia, 78 in

the case of Mexico, 28 in the case of France, 22 for Ireland, 9 for Portugal and 56 in the case of

3 Germany is one the strongest markets in Europe and is the one of the largest in the world, thus, it is reasonable to

use this country as a benchmark for the European economies. Meanwhile, despite the differences in size and level of

development, the proximity and economic ties between Latin American economies and the United States make this

country a wiser benchmark for the region.


13

Venezuela. Despite the efforts of constructing the most comprehensive panel set, data is still

highly limited for the Latin American economies; in consequence the panel compiled is uneven4.

Graph No. 1 helps describing the nature of the missing data. As it can be analyzed missing

values are present mostly in financial variables. The dark red bar in the graph represents the total

missing data for each variable out of the 2652 observations possible in the complete panel. It can

be inferred that country specific variables and the common factor – VIX – present almost full

information; unfortunately, the Bid-Ask liquidity variable is only available in 65% of the total

possible number of observations and for the yield spread this figure rises only to 84%.

Chart No. 1

The lack of information can be explained by the limitations of Latin American countries where

432 observations are missing out of the 1092 total Latin American observations possible, this

4 Appendix 1 explains in more detail the characteristics of the data collected and compiled.

0

78

24

121

12

939

432

2652

2574

2628

2531

2640

1713

2220

0 500 1000 1500 2000 2500

VIX

GDP Growth

Current Account /GDP

Gov. Debt / GDP

Private Debt /GDP

Bid Ask Spread

Yield Spread10y TR & B

Missing values Observations


14

means that 39% of the data of the dependent variable is missing for the Latin American

economies.

Furthermore, table No. 2 presents the expected signs of the estimations of equation (2) with the

information available in the uneven panel compiled, along with a brief description and the

intuition behind its consideration in the model:

Table No. 2 Variable Description and Intuition Expected Sign

Complete

Panel

Expected

Sign

LA Panel

Yield Spread (First

Lag)

Accounts for the influence of previous

period information on current yield

spreads. Captures persistence in the

yield spreads.

+

+

GDP Growth Controls for the dynamic performance

of an economy. - -

Current Account /

GDP

Controls for the ability of a country to

pay its liabilities. It is also a proxy for

the net foreign assets of a country.

- -

Gov Debt /GDP Controls for the stock of debt of the

country. Gives signals to the market

about the financial position of the

country.

+ +

Private_Debt / GDP Controls for the level of domestic

leverage of a country. + +

Bid – Ask Spread Controls for the liquidity level of the

market. A narrow gap corresponds to a

liquid market and low spread

requirements.

+ +

VIX Controls for global risk. + +

Dummy Dummy variable that takes the value of

1 after October 2009, the burst of the

Greek debt crisis, and zero else.

Attempts to capture the case of pure

contagion.

+ +

Yield_Spread(First

Lag)*D

Attempts to capture whether the

persistence of the variable changes

after the burst of the Greek Debt Crisis.

+

-

GDP_G*D Attempts to capture wake – up – call

contagion.

-

-

Current_Account/G

DP*D

Attempts to capture wake – up – call

contagion.

-

-


15

Variable Description and Intuition Expected Sign

Complete

Panel

Expected

Sign

LA Panel

Gov Debt/GDP*D Attempts to capture wake – up – call

contagion.

+

-

Private Debt/GDP*D Attempts to capture wake – up – call

contagion.

+

-

Bid – Ask Spread*D Attempts to capture wake – up – call

contagion.

-

-

VIX*D Attempts to capture shift contagion. + -

Note that the expected signs of the interacted terms with the dummy, corresponding to the Greek

debt crisis, differ between the estimation with the complete panel and the estimation with the

Latin American Panel. This divergence appears because markets started evaluating the regions

differently. In the case of the Latin American panel the behavior of sovereign bonds markets

should reflect the effects of a positive wake –up – call contagion, a tightening of yield spreads

caused by a reassessment of fundamentals. The latter would be represented by negative

coefficients of the interacted terms between the fundamentals and the dummy crisis variable.

Meanwhile, and in addition to the data availability obstacle presented for the Latin American

region previously discussed, the complete panel presents a slight over representation of European

economies – 10 European vs. 7 Latin American economies. In consequence, the coefficients of

the interacted terms are expected to be dragged, to some extent, by the behavior of the European

markets showing the traditional case of wake – up –call contagion. Hence, coefficients of the

interacted terms in the complete panel should reaffirm and deepen the expected signs of the

variables with no interaction.

Results

Table No. 3 below presents the results for the estimations based on the complete panel (Model 1)

and the Latin American panel (Model 2). These estimations are based on 1604 observations for

the complete panel and 170 observations for the Latin American Panel, as only lines with

complete information for all the variables were included in the estimation. Argentina for example

was excluded from the analysis since it does not have any reporting data for the Bid – Ask spread

variable.


16

Table No. 3

Dependent variable: Yield Spread Variable Model 1 Model 2 Variable Model 1 Model 2

HAC - FE HAC - FE

Constant .0000 (.0005)

.0268

(.0120) *

Yield Spread

(Lag1) .9014

(31.11) ***

.4306

(.0463)

***

Yield

Spread(lag1)*

D

.0323 (.0250)

-.2384

(.0694) **

GDP Growth -.00782

(.0062)

.0028

(.0036) GDP

Growth*D -.0010 .0089

-.0014

(.0097)

Current

Account % GDP .0074

(.0036) *

.1009

(.0647) Current

Account %

GDP*D

-.0261 .0042 ***

-.1411

(.0678) *

Gov Debt %

GDP .0011

(.0019)

.0330

(.0048)

***

Gov Debt %

GDP*D .0019

(.0008) **

-.0573

(.0072) ***

Private Debt %

GDP .0004

(.0007)

.0062

(.0024) *

Private Debt

% GDP*D -.0004 .0006

-.0082

(.0029) **

Bid-Ask Spread .0032 (.0020)

.0091

(.0016) ***

Bid-Ask

Spread*D -.0033

(.0020) -.0091

(.0016) ***

VIX .0049 (.0013)

***

.0647

(.0455) VIX*D .0034

(.0070)

-.0432 .0444

D -.0019 .0019

-.0098

(.0143)

No. observations

1604

170

R-squared .9704 .4699 R-squared adj .9382 .7388 Prob(F-statistic) .0000 - Rho coefficient .5072 .9688

Standard errors in parenthesis; *** Significant at 1%, ** Significant at 5%, * Significant at 10%

A Hausman test was conducted in order to confirm that the correct model is in fact a fixed effects model rather than

a random effects model. A Wald test for heteroskedasticity and a Wooldridge test for autocorrelation were also

conducted; they confirm the need to correct for the presence of heteroskedasticity and autocorrelation


17

The foremost finding lays in the results of Model 2 which provide evidence of positive wake – up

– call contagion towards the Latin American region. This can be inferred from the significant

and negative sign of the coefficients of the interacted terms between the crisis dummy and the

current account to GDP ratio, the government debt to GDP ratio, the private debt to GDP ratio

and the Bid-Ask spread. Additionally, and in line with the results reported in Giordano et al.

(2012), wake – up – call contagion is also found in the results for the complete panel – Model 1.

Furthermore, in line with the results presented in Giordano et al.(2012), the estimations do not

provide evidence of pure contagion nor shift contagion.

An interesting finding in Model 2 is that the common factor is not significant. This suggests that

Latin American countries were isolated from international financial markets, probably because

the region was deleveraged since the economic and financial crisis experienced during the 90s.

Another interesting finding corresponds to the magnitudes of the coefficients in Model 2. First of

all, note that this model presents significant coefficients and expected signs for the government

to GDP ratio, the private debt to GDP ratio and the Bid-Ask spread. These results indicate that

markets evaluated the LA region according to their fundamentals even before the crisis.

Secondly, note that partial elasticities between yield spreads and each of the fundamentals

provide a negative result5. This discovery indicates that after the Greek debt crisis markets

considered that yield spreads in the LA merited a net decrease.

Results of Model 1 present high persistence of the dependent variable, a strong sensitivity to

common shocks and a high sensitivity to fundamentals only after the Greek debt crisis6. The

signs of the significant coefficients are the expected, except for the current account to GDP ratio

which has small but positive sign; nevertheless, the wake – up - call contagion effects corrects

the effect and gives a negative partial elasticity for the yield spreads with respect to this variable.

5 Partial elasticities should be considered as follows: δyield_spread/δ”fundamental” │D=1.

6 Similar result were obtained by Arghyrou and Kontonikas (2012), in which a marked shift after the European debt

crisis from a convergence trade model to a macro-fundamentals and international risk driven model is found.


18

Moreover, results show that the lagged variable is strongly significant in all estimations, but the

marginal effect of the crisis on this variable is zero in Model 1 and negative in Model 2. This

means that in Model 1 the crisis did not change the perception of the markets regarding historic

information, but in Model 2, for Latin American countries the persistence of the variable was

lessened. Finally, in Model 1 the coefficient of the lagged dependent variable reaches the 0.9

levels. Given this outcome and the possibility of having a non-stationary yield spread variable, a

unit root test is conducted using the Maddala-Wu (1999) and the Im, Shin and Pesaran (2003)

approaches. Results show that the null hypothesis of unit root can be rejected for both models.

Stationarity does not seem to be an obstacle in the estimations.

Robustness checks

Five robustness checks were also performed in order to check the stability of the results

presented in Table No. 3.

The first and second robustness checks comprise the use of the second and third lag as

instrumental variables (IV) for the lagged yield spread, in order to avoid some correlation

existing when using the first lag as explanatory variable. Results for Model 1 regarding wake-up-

call contagion - the coefficients of the interacted terms between the fundamentals and the

dummy crisis variable - remain the same but the coefficient of the interacted private debt to GDP

ratio is now significant. Nevertheless and in contrast to the results of the base model (HAC-FE),

the dummy variable shows a significant coefficient with an unexpected negative sign, indicating

that the crisis tightened yield spreads. This result could be reflecting the fact that by October

2009 yield spreads were already significantly wide due to the subprime crisis of 2008, instead

since that month yield spreads slightly tightened due to Government interventions that started to

take place since that date (Caceres, Guzzo and Segoviano (2010)). In the case of the Latin

American panel the third lag was not a good IV estimator7, thus, the second lag was the only IV

correction used. Results also show significant coefficients for the dummy variable, and for the

interacted terms with the government debt to GDP ratio, the Bid – Ask spread and the VIX.

7 The third lag appear not be significant as an explanatory variable of the yield spread variable.


19

These results suggest the existence of the three types of contagion in the Latin American region,

a result not obtained in the base model.

The third robustness check considered a shorter period of time and takes out the first

observations in which most of the missing values were present. With observations

comprehending the period between March 2006 and December 2012, estimations were

conducted and conclusions over Model 1 remain unchanged, but some additional fundamentals

appear to be significant: private debt to GDP ratio, Bid-Ask spread and its interacted term.

Meanwhile, the Latin American panel shows again evidence of the three types of contagion.

The fourth robustness check intends to capture the effects of the subprime crisis of 2008. This

led to a global financial crisis that affected heavily European markets, as well as the emerging

world. For this purpose, the dummy variable was changed in order to take the value of 1 since

September 2008, date of the bankruptcy of Lehman Brothers (BIS Papers No. 54, December

2010). Results are also very stable regarding the three types of contagion and compared to the

original results (base model). However, the current account to GDP ratio loses significance in the

complete panel, while the interacted term of the current account to GDP ratio with the dummy

loses its significance in the Latin American panel estimation.

The previous robustness checks show that results regarding the existence of a positive wake-up-

call contagion – the coefficients of the interacted terms between the dummy crisis and

fundamentals – in the Latin American region are very stable. On the other hand, results about the

existence of pure contagion – the coefficient of the dummy variable - and shift contagion – the

coefficients of the interacted term between the dummy crisis and the VIX – are sensible to

instrumental variables corrections.


20

Conclusions

The present thesis explores a possible explanation to the recent sovereign credit rating upgrades

experienced by Latin American economies. This consists in a positive wake – up – call contagion

triggered by the 2009 Greek debt crisis. The hypothesis was inspired by the work of Giordano et

al. (2012), a document that proposes the analysis of three types of contagion – (1) wake – up –

call contagion, (2) shift contagion and (3) pure contagion - experienced within the Euro zone

and caused by the Greek debt outbreak. This way, the present document follows their proposed

methodology, extending the original sample of countries to add the major Latin American

economies. Thus, I set out to analyze contagion with a positive connotation in regions such as the

Latin American. Results provide support for an episode of positive wake – up – call contagion

within the Latin American region, and indeed the Greek debt crisis provoked a tightening in

Latin American yield spreads via a reassessment of country specific variables. This confirms that

pricing of Latin American sovereign bonds in international financial markets changed.

Nowadays, international financial markets are acknowledging the efforts of Latin American

economies in correcting fiscal imbalances, implementing inflation targeting regimes and creating

capital buffers to counter the effects of new coming local and international crises8. I believe these

results are most valuable for fund managers and other fixed income investors. They suggest that

Latin American sovereign bonds represent a new trendy investment choice that provides an

option for risk diversification purposes. Results are also relevant for Latin American policy

makers since they show that strict economic efforts are being priced in international financial

markets. Findings of positive wake – up – call contagion are robust, in contrast to findings of

pure and shift contagion that were not stable in the robustness checks conducted. Nonetheless, a

warn is raise regarding the scarce availability of data in the Latin American panel, in which

estimations are conducted over 170 observations out of a total of 1092 observations possible in

the time period selected. Further analysis on the topic can consider a time varying coefficients

methodology, as proposed in Bernoth & Erdogan (2010); it can also consider enlarging the

sample of countries in order to analyse the behaviour of the rest of the emerging markets’ yield

8 Results about the Latin American region refer to Brazil, Chile, Colombia, Mexico, Peru and Venezuela. Argentina

was excluded from the analysis due to lack of information available.


21

spreads, for example from Asian economies which have shared past episodes of crisis with Latin

America and have also built stronger economies since.

V. Bibliography

Arghyrou, M. G., & Tsoukalas, J. D. (2011). The Greek debt crisis: Likely causes, mechanics

and outcomes. The World Economy, 34(2), 173-191.

Baldacci, E., Mati, A., and Gupta, S. (2008). Is it (still) mostly fiscal? Determinants of sovereign

spreads in emerging markets (No. 2008-2259). International Monetary Fund.

Baker and Mckenzie (2010). “Brazil takes off”, March 2013,

http://www.bakermckenzie.com/files/Uploads/Documents/Locations/Latin%20America/pn_latax

2010_04_braziltakesoff_mar10.pdf (downloaded 13th August 2013)

Bellas, D., Papaioannou, M., and Petrova, I. (2010). Determinants of emerging market sovereign

bond spreads: fundamentals vs financial stress. IMF Working Papers, 1-25.

Bumachar, J., and Goldfajn, I. (2012). “Latin America during the crisis: the role of

fundamentals”, Ita’u Unibanco.

Bernoth, K., & Erdogan, B. (2012). Sovereign bond yield spreads: A time-varying coefficient

approach. Journal of International Money and Finance, 31(3), 639-656.

Bases, Daniel (2013), “UPDATE 3-S&P revises Mexico sovereign credit outlook to positive”,

http://www.reuters.com/article/2013/03/12/mexico-standardandpoors-outlook-

idUSL1N0C4B7M20130312 (downloaded 13th August 2013)

Caceres, C., Guzzo, V., & Segoviano Basurto, M. (2010). Sovereign spreads: Global risk

aversion, contagion or fundamentals?. IMF working papers, 1-29.

Cárdenas, M,. & Henao, C. (2010). “Latin America and the Caribbean Economic Recovery”, The

Brookings Institution.

Constâncio, V. (2012). Contagion and the European debt crisis. Financial Stability Review, 16,

109-121.

Ebner, A. (2009). An empirical analysis on the determinants of CEE government bond spreads.

Emerging Markets Review, 10(2), 97-121.

Edwards, S. (1986). The pricing of bonds and bank loans in international markets: An empirical

analysis of developing countries' foreign borrowing. European Economic Review 30.3, 565-589.


22

Eichengreen, B., & Mody, A. (1998). What explains changing spreads on emerging-market debt:

fundamentals or market sentiment? (No. w6408). National Bureau of Economic Research.

Eichengreen, B., Mody, A., Nedeljkovic, M., & Sarno, L. (2012). How the subprime crisis went

global: Evidence from bank credit default swap spreads. Journal of International Money and

Finance, 31(5), 1299-1318.

Forbes, K., and R. Rigobon. (2000). “Contagion in Latin America: Definitions, measurement and

policy implications”, NBER Working Paper No. 7885.

Galindo, A., Izquierdo, A. & Rojas-Suarez, L. (2010). “Financial Integration and Foreign Banks

in Latin America: How do they Impact the Transmission of External Financial Shocks”

(forthcoming) CGD Working Paper. Center for Global Development, Washington, DC.

Giordano, R., Pericoli, M., & Tommasino, P. (2012).”'Pure or Wake-Up-Call Contagion?

Another Look at the EMU Sovereign Debt Crisis”, Midwest Finance Association 2013 Annual

Meeting Paper.

Goldberg, L. 2005. “The International Exposure of U.S. Banks.” NBER Working Paper 11365.

Cambridge, United States: National Bureau of Economic Research.

Im, K. S., Pesaran, M. H., & Shin, Y. (2003). “Testing for unit roots in heterogeneous panels”,

Journal of econometrics, 115(1), 53-74.

Jara, A., Moreno, R., & Tovar, C. (2009). “The global crisis and Latin America: financial impact

and policy responses”, BIS Quarterly Review, 65.

Jaramillo, Andrea (2013), “Colombia’s S&P Rating Raised on Economic Growth, Peace Talks”,

http://www.bloomberg.com/news/2013-04-24/colombia-rating-raised-by-s-p-on-economic-

expansion-peace-talks (downloaded 13th August 2013)

Kouretas, G., & Prodromos, V. (2010). “The Greek crisis: Causes and implications”,

Panoeconomicus, vol. 57, br. 4, str. 391-404

MacKinnon, J. G. (1991). Critical values for cointegration tests, Chapter 13 in Long-Run

Economic Relationships: Readings in Cointegration, Oxford University Press.

MacKinnon, J. G. (2010). Critical values for cointegration tests (No. 1227). Queen's Economics

Department Working Paper.

Maddala, G. S., & Wu, S. (1999). A comparative study of unit root tests with panel data and a

new simple test. Oxford Bulletin of Economics and statistics, 61(S1), 631-652.

Martinez, M., & Ramirez, M. (2011). “International propagation of shocks: an evaluation of

contagion effects for some Latin American countries” Macroeconomics and Finance in

Emerging Market Economies, 4:2, 213-233


23

Min, H. G. (1998). Determinants of emerging market bond spread: do economic fundamentals

matter? World Bank Publications. No. 1899.

Mody, A. (2009). From Bear Stearns to Anglo Irish: how eurozone sovereign spreads related to

financial sector vulnerability. IMF Working Papers, 1-39.

Mody, A. and Sandri, D. (2012), The Eurozone crisis: how banks and sovereigns came to be

joined at the hip. Economic Policy, 27: 199–230.

Montoro, C., & Rojas-Suarez, L. (2012). “Credit at times of stress: Latin American lessons from

the global financial crisis”, BIS Working Paper No. 370

Ocampo, J. A. (2009). Latin America and the global financial crisis. Cambridge journal of

economics, 33(4), 703-724.

QMS Quantitative Micro Software, EViews 7 User’s Guide II, Pag. 617 - 646

Reinhart, C. M., Rogoff, K. S., & Savastano, M. A. (2003). “Debt intolerance”, National Bureau

of Economic Research, No. w9908

Reinhart, C., & Rogoff, K. (2011). "From Financial Crash to Debt Crisis," American Economic

Review, American Economic Association, vol. 101(5), pages 1676-1706

Scotiabank (2013). “Latin America. Regional Outlook”, April 2013, www.scotiabank.com

(downloaded 13th August 2013)

Sgherri, S., & Zoli, E. (2009). Euro area sovereign risk during the crisis. IMF Working Papers, 1-

23.


24

Appendix

No. 1 Data Collection Details

The data collection was a challenging exercise given the uneven availability of data. For the

Latin American countries variables are not always reported quarterly, in some cases they are not

reported at all. The presence of these economies in international financial markets is also limited;

that limitation is reflected on the scarce financial information available. The following charts

describe in detail the data found and used for the purpose of this research.

Yield Spread

European Countries Latin American Countries

Source Thomson Reuters (TR) – Data

Stream

Thomson Reuters (TR) – Data Stream

Variable found TR Government Benchmark Bid

Yield 10Y

TR Government Benchmark Bid Yield

10Y

Periodicity Monthly Monthly

Transformations Missing data from TR was completed

with data found in Bloomberg.

Missing values 0 / 2652 432 / 1092

Other details Information for Perú is not found in

TR, Bloomberg was used instead.

Bid – Ask Spread


Source Bloomberg Bloomberg

Variable found BID Price – Ask Price of the 10Y

monthly reference sovereign bond for

each country

BID Price – Ask Price of the 10Y

monthly reference sovereign bond

for each country

Periodicity Monthly Monthly

Transformations Spread calculated with respect to the

corresponding benchmark, Germany

for European economies and US for

Latin America

Spread calculated with respect to the

corresponding benchmark, Germany

for European economies and US for

Latin America

Missing values 63 / 2652 876 / 1092

Other details Information for Argentina was not

found


25

GDP


Source OECD IADB

Variable found Total, current prices_millions of

national currency

Total, current prices_millions USD

Periodicity Quarterly Quarterly

Transformations Conversion from EUR to USD.

Exchange rate available at European

Central Bank.

Missing values 0 / 2652 0/1092

Other details Venezuela reports annual data. It

was left constant within the months.

* USA GDP was taken from OECD data base

GDP growth


Source OECD IADB

Variable found Growth rate compared to same

quarter previous year

Total, current prices_millions USD


Transformations Calculated growth rate with respect

to the same quarter in the previous

year.

Missing values 57 / 2652 21 / 1092

Other details Venezuela reports annual data. It

was left constant within the months.

* USA GDP was taken from OECD data base

Current Account % GDP


Source OECD IADB

Variable found Current Account Balance: % of GDP Current Account Balance: % of

GDP


Transformations

Missing values 21 / 2652 3 / 1092

* USA current account was taken from OECD data base

Government Debt % GDP


Source Eurostat IADB, OECD


26

Government Debt % GDP


Variable found Government consolidated gross debt Total Public Debt: % of GDP

Periodicity Quarterly Monthly, Quarterly, Annually

Transformations Divided the variable to GDP current

prices

Missing values 9 / 2652 112 / 1092

* USA Government Debt was taken from OECD

Private Debt % GDP


Source Eurostat IADB

Variable found Private debt in % of GDP - non

consolidated - annual data*

Credit to the Private Sector:

millions of U$S- end of period

Periodicity Quarterly data Monthly

Transformations Divided the variable by GDP

current US Dollars.

Missing values 12 / 2652 0 / 1092

** USA private debt was taken from The World Bank

* The private sector debt is the stock of liabilities held by the sectors Non-Financial

corporations (S.11) and Households and Non-Profit institutions serving households (S.14_S.15).

The instruments that are taken into account to compile private sector debt are Securities other

than shares (F.3) and Loans (F.4), that is, no other instruments are added to calculate the

private sector debt. (Eurostats http://epp.eurostat.ec.europa.eu/tgm/web/table/description.jsp ).

External Debt on Bonds and Notes / Total Gross External Debt


Source The World Bank

Variable found External Debt on Bonds and Notes and Total Gross External Debt

Periodicity Quarterly

Transformations Division of the variables found

Missing values 432 / 1092 for Latin America

Other details No data is available for Venezuela

VIX

European Countries & Latin American Countries

Source Bloomberg

Variable found VIX Index

Periodicity Monthly

Transformations No

Missing values 0


27

No. 2. Graphs and Descriptive Statistics

Graph No. 1

Yield Spreads in Latin America - 10Y Government Bonds Yield Spreads

*10Y Government Reference Bond Yield Spreads with respect to the U.S.

Source: Bloomberg and Thomson Reuters Index

-.1

.0

.1

.2

.3

.4

.5

.6

.7

2000:1 2002:1 2004:1 2006:1 2008:1 2010:1 2012:1

ARGENTINA AUSTRIA BELGIUM

BRAZIL CHILE COLOMBIA

FINLAND FRANCE GREECE

IRELAND ITALY MEXICO

NETHERLANDS PERU PORTUGAL

SPAIN VENEZUELA


28

Graph No. 2

10Y Government Bonds Yield Spreads

Graph No. 3

10Y Government Bonds Yield Spreads - Latin American Economies

-.2

.0

.2

.4

.6

.8

2000 2002 2004 2006 2008 2010 2012

ARGENTINA

-.004

.000

.004

.008

.012

.016

2000 2002 2004 2006 2008 2010 2012

AUSTRIA

-.01

.00

.01

.02

.03

2000 2002 2004 2006 2008 2010 2012

BELGIUM

.04

.06

.08

.10

.12

.14

.16

2000 2002 2004 2006 2008 2010 2012

BRAZIL

.00

.01

.02

.03

.04

.05

2000 2002 2004 2006 2008 2010 2012

CHILE

.00

.04

.08

.12

.16

2000 2002 2004 2006 2008 2010 2012

COLOMBIA

-.002

.000

.002

.004

.006

.008

2000 2002 2004 2006 2008 2010 2012

FINLAND

.00

.01

.02

.03

.04

2000 2002 2004 2006 2008 2010 2012

FRANCE

.0

.1

.2

.3

.4

2000 2002 2004 2006 2008 2010 2012

GREECE

-.02

.00

.02

.04

.06

.08

.10

2000 2002 2004 2006 2008 2010 2012

IRELAND

.00

.01

.02

.03

.04

.05

.06

2000 2002 2004 2006 2008 2010 2012

ITALY

.02

.03

.04

.05

.06

.07

2000 2002 2004 2006 2008 2010 2012

MEXICO

-.002

.000

.002

.004

.006

.008

.010

2000 2002 2004 2006 2008 2010 2012

NETHERLANDS

.008

.012

.016

.020

.024

.028

.032

2000 2002 2004 2006 2008 2010 2012

PERU

-.04

.00

.04

.08

.12

.16

2000 2002 2004 2006 2008 2010 2012

PORTUGAL

-.02

.00

.02

.04

.06

2000 2002 2004 2006 2008 2010 2012

SPAIN

-.05

.00

.05

.10

.15

.20

2000 2002 2004 2006 2008 2010 2012

VENEZUELA

-.2

.0

.2

.4

.6

.8

00 01 02 03 04 05 06 07 08 09 10 11 12

ARGENTINA

.04

.06

.08

.10

.12

.14

.16

00 01 02 03 04 05 06 07 08 09 10 11 12

BRAZIL

.00

.01

.02

.03

.04

.05

00 01 02 03 04 05 06 07 08 09 10 11 12

CHILE

.02

.04

.06

.08

.10

.12

.14

00 01 02 03 04 05 06 07 08 09 10 11 12

COLOMBIA

.02

.03

.04

.05

.06

.07

00 01 02 03 04 05 06 07 08 09 10 11 12

MEXICO

.008

.012

.016

.020

.024

.028

.032

00 01 02 03 04 05 06 07 08 09 10 11 12

PERU

-.04

.00

.04

.08

.12

.16

.20

00 01 02 03 04 05 06 07 08 09 10 11 12

VENEZUELA


29

Descriptive Statistics - Complete Panel

Table 4.

YIELD

SPREAD GDP Growth Current

Account % GDP

Gov. Debt % GDP

Private Debt % GDP

BID ASK SPREAD

VIX

Mean 0.0258 0.0350 -0.0110 0.1435 0.0157 -0.4450 0.2171

Median 0.0041 0.0115 0.0030 0.1280 0.0260 -0.0850 0.2027

Maximum 0.6227 0.4884 0.2419 1.3020 2.7884 0.0000 0.5989

Minimum -0.0206 -0.6893 -0.2430 -0.7903 -1.5681 -87.3490 0.1042

Std. Dev. 0.0470 0.1077 0.0790 0.4025 0.8138 3.0685 0.0830

Skewness 4.4939 -0.2427 -0.0993 0.0524 0.0616 -18.4148 1.5489

Kurtosis 37.639 12.1690 3.3265 2.5554 2.5508 433.1906 6.7175

Jarque-Bera 118460.8 9042.025 16.0057 22.006 23.8634 13305753 2587.570

Probability 0.0000 0.0000 0.0003 0.0000 0.0000 0.0000 0.0000

Sum 57.442 90.3380 -29.0466 363.4191 41.5389 -762.3870 575.9719

Sum Sq. Dev. 4.9054 29.8700 16.4175 410.0777 1747.876 16120.06 18.2856

Observations 2220 2574 2628 2531 2640 1713 2652

Descriptive Statistics - Panel Latin America

Table 5.

YIELD SPREAD

10Y

GDP Growth Current Account %

GDP

Gov. Debt % GDP

Private Debt % GDP

BID ASK SPREAD

VIX

Mean 0.0624 0.0772 0.0519 -0.1506 -0.5374 -2.3143 0.2171

Median 0.0497 0.0766 0.0381 -0.2049 -0.6556 -0.3880 0.2027

Maximum 0.6227 0.4884 0.2419 1.1281 2.7884 0.0000 0.5989

Minimum -0.0206 -0.6893 -0.0114 -0.7903 -1.5681 -87.349 0.1042

Std. Dev. 0.0591 0.1546 0.0486 0.3539 0.7990 8.3394 0.0830

Skewness 4.1384 -1.0008 1.6370 0.8344 1.2030 -6.6293 1.5489

Kurtosis 31.6729 7.5122 5.9687 4.2933 4.4233 57.2216 6.7175

Jarque-Bera 24492.78 1087.392 886.3169 182.0253 355.5843 28042.01 1065.470

Probability 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000

Sum 41.23526 82.71145 56.58540 -147.6729 -586.9370 -499.9060 237.1649

Sum Sq. Dev. 2.3033 25.6066 2.5716 122.6705 696.5998 14952.37 7.5293

Observations 660 1071 1089 980 1092 216 1092


30

No. 3. Hausman Test for Fixed or Random Effects

Model 1 Model 2

Chi2 63.53 138.99

Prob>chi2 0.0000 0.0000

The null hypothesis that states that the errors are not correlated with the independent variables is

rejected. Controlling for fixed effects is correct for both Model 1 and 2.

No. 4 Wald Statistic for Heteroskedasticity

Model 1 Model 2

Chi2 49169.99 861.17

Prob>chi2 0.0000 0.0000

The null hypothesis of homoskedasticity is rejected for both models at a 1% significance level.

No. 5 Wooldridge Test for Autocorrelation

Model 1 Model 2

F 62.409 28.756

Prob>F 0.0000 0.0030

The null hypothesis of non existence of serial correlation is rejected for both models at a 1%

significance level.

No. 6 Im-Pesaran-Shin Unit Root Test

Model 1

Im, Shin and

Pesaran

Model 1

Maddala

and Wu

Model

1

Model 2

Im, Shin and

Pesaran

Model 2

Maddala

and Wu

Model

2

Variable Yield Spread Yield Spread Resid Yield Spread Yield

Spread

Resid


31

Statistic -1.4501 44.9715 -1.7285 -2.1469 22.07 -3.0555

P-value

0.0735

0.0988

0.0420

0.0159

0.0771

0.0011

The results of the unit root tests show that the null hypothesis of unit root can be rejected for

both models. Stationarity does not represent an obstacle in the estimations.

In order to analyze deeper this result, a unit root test is conducted over two different sample

periods, before and after the crisis.

Unit Root Test – Before crisis period

Model 1

HAC

Model 1

HAC

Model 2 Model 2

Variable Yield Spread Resid Yield

Spread

Resid

Statistic

0.9292

not available

-0.1134

not available

P-value 0.8236

Critic. Val

1%

5%

10%

-5.84

-5.29

-5.004

-5.79

-5.25

-4.97

* Estimations of p-values for cointegration tests considered in MacKinnon (2010)

The yield spread variable in this case has a unit root. Given this result an Engle Granger

Cointegration Test is conducted but it delivers inconclusive results. A near to singular matrix

does not allow the estimation of the test over the residuals.

Unit Root Test – After crisis period

Model 1

HAC

Model 2

Var Yield Spread

Yield Spread

Statistic -3.4586 -3.8710

P-value 0.0003 0.001


32

In the case of the after crisis period, results indicate that yield spreads in the complete panel are

I(0).

No. 7 Robustness Checks

Complete Panel

Table 6.

Dependent variable: Yield Spread Variable Model 1 Model 1 Model 1 Model 1 Model 1

HAC

HAC IV: Lag2

HAC IV: Lag3

HAC Panel less

observations

HAC Dummy 2

Constant .0000 (.0005)

.0001

(.0004) .0002

(.0004) .0031

(.0011) **

.0004

.0004

Yield Spread

(First Lag) .9014

(31.11) ***

.9857

(.0209) ***

1.033

(.0243) ***

.8222

(.0403) ***

.9275 (.0277)

*** GDP

Growth -.00782

(.0062) -.0028

(.0038) .0003

(.0040) -.0105

(.0101)

-.0100 (.0071)

Current

Account %

GDP

.0074 (.0036)

*

.0083

(.0036) **

.0089

(.0037) **

.0362

(.0107)

***

.0077 (.0044)

Gov Debt %

GDP .0011

(.0019)

.0004

(.0010) -.0000

(.0011) .0015

(.0020) .0010

(.0018)

Private Debt

% GDP .0004

(.0007)

.0008

(.0005) .0010

(.0005) *

-.0015

(.0008) *

.0001 (.0008)

Bid-Ask

Spread .0032

(.0020) .0049

(.0012) ***

.0057

(.0013) ***

.0062

(.0020)

***

.0010 (.0018)

VIX .0049 (.0013)

***

.0031

(.0013) **

.0021

(.0014) .0100

(.0016)

***

.0009 (.0015)

** D -.0019

.0019 -.0015

.0009 *

-.0012

(.0009) -.0016

(.0021) -.0007 (.0009)

Yield

Spread(First

Lag)*D

.0323 (.0250)

-.0397

(.0197) **

-.0804

.0225

***

.0748

(.0397) *

.0139 (.0239)

GDP -.0010 -.0059 -.0091 .0042 -.0033


33

Variable Model 1 Model 1 Model 1 Model 1 Model 1 HAC

HAC

IV: Lag2 HAC

IV: Lag3 HAC

Panel less

observations

HAC Dummy 2

Growth*D .0089

(.0049) .0051 *

(.0102) .0102

Current

Account %

GDP*D

-.0261 .0042 ***

-.0213

(.0044) ***

-.0187

(.0045) ***

-.0326

(.0057) ***

-.0161 .0028 ***

Gov Debt %

GDP*D .0019

(.0008) **

.0014

(.0007) *

.0011

(.0007) .0022

(.0010) **

.0012 (.0005)

* Private Debt

% GDP*D -.0004 .0006

-.0009

(.0004) *

-.0011

(.0004) **

.0008

(.0006) -.0004 (.0006)

Bid-Ask

Spread*D -.0033

(.0020) -.0049

(.0012) ***

-.0058

(.0013) ***

-.0062

(.0020) ***

-.0010 .0019

VIX*D .0034 (.0070)

.0049

(.0032) .0057063

(.0032) *

-.0040

(.0079) .0018

(.0023)

No. Obs

1604

1592

1580

888

1604

R-squared 0.9704 0.9720 0.9718 0.9627 0.9718 R-squared

adjusted .9382 - - 0.9285 0.9365

Prob(F-Stat) 0.0000 0.0019 0.0378 0.0000 0.0000 Rho

coefficient

.5072

.2857

.1681

0.4254

0.3383

Latin American panel

Table 7.

Dependent variable: Yield Spread Variable Model 2 Model 2 Model 2 Model 2 Model 1

HAC

HAC IV: Lag2

HAC IV: Lag3

HAC Panel less

observations

HAC Dummy 2

Constant .0268

(.0120) *

.0447

(.0114)

***

-

.1027

(.0023)

***

.0271

(.0124) *

Yield Spread

(First Lag) .4306

(.0463)

***

.1334

(.1482) -

.6585

(.0314)

***

.3919

(.0746)

*** GDP

Growth .0028

(.0036) .0008

(.0132) -

-.0299

(.0074)

.0044

(.0027)


34


HAC

IV: Lag2 HAC

IV: Lag3 HAC

Panel less

observations

HAC Dummy 2

**

Current

Account %

GDP

.1009

(.0647) .0053

(.0832)

-

-.1516

(.0440)

**

.0974

(.0824)

Gov Debt %

GDP .0330

(.0048)

***

.0413

(.0125)

***

-

.0663

(.0075)

***

.0409

(.0193) *

Private Debt

% GDP .0062

(.0024) *

.0008

(.0042)

-

.0401

(.0015)

***

.0070

(.0036)

Bid-Ask

Spread .0091

(.0016)

***

.0123

(.0038)

***

-

.0085

(.0004)

***

.0077

(.0033) *

VIX .0647

(.0455) .0873

(.0262)

***

-

-.0096

(.0107) .0685

.0446

D -.0098

(.0143) -.0338

(.0176) *

-

-.0825

(.0058)

***

-.01566

(.01475)

Yield Spread

(First

Lag)*D

-.2384

(.0694) **

-.0727

(.1105)

-

-.4956

(.0587)

***

-.1543

(.1172)

GDP

Growth*D -.0014

(.0097) .0047

(.0159)

-

.03306

(.0085)

**

-.0046

(.0074)

Current

Account %

GDP*D

-.1411

(.0678) *

-.0935

(.0967)

-

.1192

(.0527) *

-.1150

(.0704)

Gov Debt %

GDP*D -.0573

(.0072)

***

-.0891

.(0211)

***

-

-.0995

(.0094)

***

-.0724

(.0174)

***

Private Debt

% GDP*D -.0082

(.0029) **

-.0040

(.0038)

-

-.04218

(.0016) ***

-.0093

(.0028)

**

Bid-Ask

Spread*D -.0091

(.0016) ***

-.01247

(.0038)

***

-

-.0086

(.0004) ***

-.0078

(.0033) *

VIX*D -.0432

.04444

-.0659

(.0294) **

-

.0329

(.0131) *

-.0512

(.0435)

No. Obs 170 165 - 127 170

R-squared 0.4699 0.0595 - 0.3519 0.5283 R-squared

adjusted 0.7388 - - 0.215 0.7180


35


HAC

IV: Lag2 HAC

IV: Lag3 HAC

Panel less

observations

HAC Dummy 2

Prob(F-stat) - 0.0000 - - - Rho

coefficient 0.9688 0.9769

- 0.9722 0.9639

* IV Yield Spread Lag 3: The third lag of the Yield Spread variable is not an accurate IV

estimator according to its test of significance as a predictor of the 1 period lagged Yield Spread.